QwenLM
Any plan for Qwen2 14B and Qwen2-32B?
I will stick to Qwen1.5-14B-Chat-GPTQ-Int4, if Qwen2 14B is not available
这里做下解释。在Qwen2本次设计中，MoE是medium size的模型主力，仅仅激活14B，但是达到30B模型的效果。但当前生态对MoE的支持还不完善，57B的模型大小对于显存要求较高，我们正在计划补上14和32这两个size的模型，但这两个模型毕竟是比较大的模型，还需要一些时间。
We previously hope that the MoE model can be your choice for a medium-size model. It actually only activates only 14B params in each forward pass but it can achieve the expected performance of a around 30B dense model. We are at the same time pushing the support for this model in the community and we understand that this 57B model requires large RAM. We are planning about providing 14B and 32B models, but this should take some time. Stay tuned.
这里做下解释。在Qwen2本次设计中，MoE是medium size的模型主力，仅仅激活14B，但是达到30B模型的效果。但当前生态对MoE的支持还不完善，57B的模型大小对于显存要求较高，我们正在计划补上14和32这两个size的模型，但这两个模型毕竟是比较大的模型，还需要一些时间。
We previously hope that the MoE model can be your choice for a medium-size model. It actually only activates only 14B params in each forward pass but it can achieve the expected performance of a around 30B dense model. We are at the same time pushing the support for this model in the community and we understand that this 57B model requires large RAM. We are planning about providing 14B and 32B models, but this should take some time. Stay tuned.
预计大概什么时候发布啊？
